| Mangroves are a special type of wetland.The uniqueness of its living environment enables the mangrove ecosystem to possess a considerable amount of material energy,which also increases the difficulty level of field investigation,leading to the limitations of mangrove investigation.As a new technology,lidar can quickly and efficiently obtain forest information and provide a new option for mangrove investigation.In order to explore the feasibility of Li Dar estimation of 3d information of mangrove,this study combined with field measurement data,separated the obtained mangrove Li Dar point cloud into single wood by different point cloud segmentation methods,and further studied the extraction of structural parameters of mangrove.The main research results of the mangrove forest in dongzhai harbor are as follows:（1）point cloud normalization is a key step for single wood segmentation and structural parameter estimation.In this study,ground points are directly used to obtain the normalized point cloud.After comparative analysis,it is found that compared with the traditional method,the normalized point cloud based on ground points has better effect,and the difference between the obtained maximum elevation value and the maximum tree height truth value is smaller.（2）study the influence of interpolation method and point cloud density on the quality of generated CHM.Three interpolation methods were used to construct CHM.After comparative analysis,it was found that from the perspective of average CHM height,IDW method produced CHM with the best quality.CHM is generated from point cloud data of different densities,and the results show that higher quality CHM can be obtained by reducing point cloud density appropriately.There is a certain relationship between the interpolation method and the point cloud density and the mass of CHM.（3）using different segmentation methods,single wood of mangrove point cloud was segmented with point cloud data of the same place and different densities,and three evaluation indexes,namely f-score（F）,Precision（P）and Recall（R）,were used for evaluation.The results show that in the study area,the optimal point cloud density based on layer-stack seed point segmentation is 30 points/m~2,the detection rate R is 0.19-0.76,the correct segmentation rate P varies between 0.38 and 0.73,and the overall accuracy F varies from 0.26to 0.73.The optimal density based on point cloud segmentation is 10 points/m~2,the detection rate R fluctuates between 0.16 and 0.41,the correct segmentation rate P is above 0.64,and the overall accuracy F is between 0.27 and 0.55.The optimal point cloud density based on CHM segmentation is 30 points/m~2,the detection rate R is between 0.11 and 0.47,and the correct segmentation rate is greater than 0.67.The minimum value of the overall precision F is 0.2,and the maximum value is 0.57.（4）based on the optimal segmentation combination,the mangrove was segmented,and the single wood parameter that was correctly segmented in the segmentation result was combined with the plot data to explore the relationship between the mangrove information obtained by Li Dar point cloud and the field measurement data.The results show that the structure parameters obtained from Li Dar data have a certain correlation with the measured values,among which the tree height has a high correlation of 0.35.The correlation of crown width was low,the correlation of crown width was high,and the correlation coefficient of red sea olive was as high as 0.95.At the same time,the linear correlation between estimating tree height and estimating crown width and actual DBH is weak.In summary,it is feasible to obtain the parameters of single wood structure of mangrove forest by lidar.This study explored the influence of different cloud density and different segmentation methods on the single wood parameters of mangrove extracted by lidar,provided an efficient technical scheme,and provided practical basis for the application of lidar in mangrove investigation. |
PDF Full Text Request